PixToPDF
VI EN

OCR PDF and images to text

Recognize text in scanned PDFs, photos, JPG, PNG and WebP files, then download it as TXT. Runs entirely in your browser — no upload, no server-side storage.

Choose or drop PDF / images Scanned PDF, JPG, PNG, WebP · up to 20 pages per run · no upload
Choosing the right language reduces errors. Vietnamese and English work offline; other languages download a model when selected.
Applies to each PDF; leave empty for all pages (up to 20 pages per run).

  

OCR does not reconstruct layout, tables or image positions; the result is reading-order text. For PDFs that already contain a text layer, use the PDF-to-text tool for faster and more accurate extraction. Tesseract.js, its core and Vietnamese/English data are bundled in /assets/tesseract/; other languages download once from tessdata.projectnaptha.com. Your files are never uploaded.

Frequently asked questions

Does OCR run entirely in the browser?

Yes. PDFs are rendered with pdf.js, then tesseract.js recognizes text on each page inside the browser. No request contains file content; you can disconnect from the network after the page and language data have loaded.

Are files uploaded to a server?

No. Files are read with the browser File API only. PixToPDF has no file-upload endpoint for OCR and does not store results on a server.

Which languages and formats are supported?

OCR supports Tesseract languages including Vietnamese, English, Chinese, Japanese, Korean, French, German, Spanish and many more. Vietnamese and English are bundled; other languages download once when selected. Inputs are PDF (including scans) and JPG, PNG or WebP images; output is a TXT file.

How many pages can one OCR run process?

Up to 20 pages per run, 100 MB per file and 300 MB total to protect memory. For longer PDFs, split them first; use sharp, high-resolution photos for better accuracy.